Pluggable Agentic AI Framework is a hands-on workshop that builds a multi-agent customer support system on AWS using Amazon Bedrock AgentCore. It follows the scenario of AnyCompany, an online retailer, where a supervisor orchestrator dynamically discovers and delegates to specialized agents via harness

Hands-on workshop to build a production AI agent with Strands Agents, tools, hooks, memory, multi-agent, evals, and deploy to Amazon Bedrock AgentCore.

Enterprise Supply Chain AI Assistant powered by Amazon Bedrock AgentCore & Strands Framework. Features secure VPC isolation, OpenSearch RAG, and automated CDK deployment.

An enterprise-grade, omnichannel voice assistant for QSRs. Built with 100% serverless AWS architecture, leveraging Amazon Bedrock AgentCore, Nova Sonic v2, and the MCP protocol for ultra-low latency, real-time voice interactions.

Open-source memory manager daemon for AI agents. Filesystem-native, LLM-pluggable, framework-agnostic. Works with OpenClaw, Strands, LangChain, AgentCore Runtime or any agent that can write a file.

A system that generates recorded browser session videos from public webpages, based on a webpage URL and the description of the needed clickthrough recording.

Reference repository for deploying and integrating Strands Agents in TypeScript with Amazon Bedrock AgentCore Runtime.

AI incident assistant: a supervisor delegates to specialist agents (observability + web context) on AgentCore and Strands with MCP — fully deployable via AWS CDK.

This code demonstrates building production-grade agentic applications with Strands Agents SDK, Amazon Bedrock AgentCore Memory, and AgentCore Runtime — from local prototyping through to cloud-hosted deployment.

Enterprise identity on Amazon Bedrock AgentCore with Lark (Feishu) as the IdP. A Gateway Request Interceptor forwards the user's identity and injects a per-user credential — the agent never holds it — so downstream MCP tools act as the user against Lark and reach only what that user can.

Transform unstructured media contracts into searchable metadata using multi-modal AI. Extracts territory grants, compensation terms, expiration dates, and clause relationships from rights deals, licenses, and talent agreements at scale.

Production-grade AI agent security with AWS Bedrock AgentCore, Cedar authorization, Cognito authentication, LangGraph, and Langfuse observability.

Minimal Amazon Bedrock AgentCore runtime preinstalling Claude Code on Node 24 with persistent session storage; pure SST v4/Pulumi infra (no agentcore CDK), least-privilege IAM, interactive SigV4 shell.

Chat web application for strands-compose

Enterprise identity on Amazon Bedrock AgentCore with Lark (Feishu) as the IdP. Per-user Lark tokens live in the AgentCore Identity Token Vault (OAuth 3LO) — stored, refreshed, and injected natively, no custom interceptor. Downstream MCP tools act as the user against Lark, reaching only what that user can.

A dating-sim-style speaking game with Shizuku and Chitose as voice companions. Uses Amazon Nova 2 Sonic for realtime voice play and a separate hidden runtime for turn analysis. The player clears staged scenes by speaking naturally, keeping the conversation romantic, fluent, and on-topic.